Contrastive Syn-to-Real Generalization

Authors: Wuyang Chen, Zhiding Yu, Shalini De Mello, Sifei Liu, Jose M. Alvarez, Zhangyang Wang, Anima Anandkumar

ICLR 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the effectiveness of CSG on various synthetic training tasks, exhibiting state-of-the-art performance on zero-shot domain generalization.
Researcher Affiliation Collaboration 1The University of Texas at Austin 2NVIDIA 3California Institute of Technology
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code Yes Codes is available at https://github.com/ NVlabs/CSG.
Open Datasets Yes The Vis DA-17 dataset (Peng et al., 2017) provides three subsets (domains), each with the same 12 object categories. ... Cityscapes (Cordts et al., 2016) contains urban street images taken on a vehicle from some European cities. ... GTA5 (Richter et al., 2016) is a vehicle-egocentric image dataset collected in a computer game with pixel-wise semantic labels.
Dataset Splits Yes For Vis DA-17, we choose Image Net pretrained Res Net-101 (He et al., 2016) as the backbone. We fine-tune the model on the source domain with SGD optimizer of learning rate 1 10 4, weight decay 5 10 4, and momentum 0.9. Batch size is set to 32, and the model is trained for 30 epochs. λ for LNCE is set as 0.1. ... The Vis DA-17 dataset (Peng et al., 2017) provides three subsets (domains), each with the same 12 object categories. Among them, the training set (source domain) is collected from synthetic renderings of 3D models under different angles and lighting conditions, whereas the validation set (target domain) contains real images cropped from the Microsoft COCO dataset (Lin et al., 2014).
Hardware Specification No The paper does not explicitly describe the specific hardware (e.g., GPU or CPU models) used for running experiments.
Software Dependencies No The paper mentions software like PyTorch but does not provide specific version numbers for any software dependencies required to reproduce the experiments.
Experiment Setup Yes For Vis DA-17, we choose Image Net pretrained Res Net-101 (He et al., 2016) as the backbone. We fine-tune the model on the source domain with SGD optimizer of learning rate 1 10 4, weight decay 5 10 4, and momentum 0.9. Batch size is set to 32, and the model is trained for 30 epochs. λ for LNCE is set as 0.1.